Subscribe
Share
Share
Embed
FUTURE OF XYZ is a bi-weekly interview series that explores big questions about where we are as a world and where we’re going. Through candid conversations with international experts, visionary leaders and courageous changemakers- we provoke new thinking about what's coming down the pipeline on matters related to art & design, science & innovation, culture & creativity.
Future of XYZ is presented by iF Design, a respected member of the international design community and host of the prestigious iF DESIGN AWARD since 1953. The show is also a proud member of the SURROUND Podcast Network. For more information, visit ifdesign.com/XYZ.
00:00:04:00 - 00:00:22:09
Speaker 1
Hi and welcome to this just before Halloween episode of Future of XYZ. We are talking about a topic today that our guest, Tey Bannerman, put together, and I'm so excited to do it today. Thanks for being on Future of XYZ.
00:00:22:11 - 00:00:24:18
Speaker 2
Thanks for having me.
00:00:24:20 - 00:00:56:08
Speaker 1
Well, you have your roots both as a product designer, but also as a software engineer, which is a very cool combination and something I think that kind of is what we're what we'll end up talking about a lot today. You spent over 20 years in the tech industry. You just recently stepped away after eight and a half years as a partner at the global management consultancy McKinsey and Company, where you had been the head of their design practice or I guess it's McKinsey Design Europe.
00:00:56:10 - 00:01:26:04
Speaker 1
As well as, like, leading and building the firm's digital and design capabilities, especially related to AI transformation. I know you have lots of other background that's really interesting. We’ll weave it into the conversation, but I guess the biggest piece of everything that I'm curious about is this topic, right? So let's start as we always do, which is like in the context of your expertise and today's conversation, how do we define being replaced?
00:01:26:06 - 00:02:06:17
Speaker 2
Easy question. No I'm kidding. I think, with, I probably define being replaced as like in two dimensions, one actually being replaced. So people's jobs, people's functions, people's roles being replaced by insert whatever here, whether it's AI, whether it's some other technology. And then I think the other side of that is around human cognition and thought processes being outsourced to technology, and what that means for long term decision making, long term ability to recover from failure, and so on.
00:02:06:19 - 00:02:30:16
Speaker 1
Wow. Okay. Well, we had joked in preparing for this conversation that the episode is perfect for this day before Halloween because it's pretty spooky to imagine being replaced. I find that there's a lot of talk about this right now, especially about jobs going away because of AI. So that first of the things that you've just defined, what's your take on that?
00:02:30:16 - 00:02:34:12
Speaker 1
And is the reality as scary as it's being made out to be?
00:02:34:14 - 00:02:52:20
Speaker 2
Yeah. I mean, I don't think the reality, as scary as some are making it out to be. I think what I'm seeing is like this dichotomy where it's either everything's really bad and awful and really scary or everything is great and AI is going to fix the world. I think, you know, as with most things, the reality is probably somewhere in between.
00:02:52:20 - 00:03:22:09
Speaker 2
But my take is that some, some functions, a lot of tasks are going to be replaced by things and, and workflows and tools that are more efficient at handling those tasks. I think a lot of jobs are going to disappear. I think a lot of new jobs are going to come in that didn't exist before. So I think things are going to change, but I don't think that the net result will be doom and gloom if things go the the right way.
00:03:22:11 - 00:03:23:13
Speaker 1
Totally fair.
00:03:23:15 - 00:03:24:17
Speaker 2
That's an if, a small if.
00:03:24:17 - 00:03:41:23
Speaker 1
That’s a small if, you know, I, I had, I had a professor on, a few weeks ago who teaches at Stanford in law and technology. That's her intersection. She's a human rights lawyer. So she's talking all about the privacy and the governance and these kinds of things. And she said, “Everything will be all right as long as we get the rights right.”
00:03:42:00 - 00:03:42:19
Speaker 1
You know, like.
00:03:42:22 - 00:03:43:17
Speaker 2
Yeah.
00:03:43:19 - 00:04:02:02
Speaker 1
It's, I butchered the quote, but it's like that. It's, it makes sense. I’ve seen some data just like the last week that was really interesting to me. And a little eye opening, and it was like the top 60 jobs the AI is going to replace and the, the, the most, at least, lowest risk jobs to be replaced.
00:04:02:03 - 00:04:27:06
Speaker 1
Yeah. And it was this crazy, like, weird world where, like, the top most replaceable jobs are these what we traditionally in recent modern capitalism have called like high skill, you know, high touch, white collar for sure. And then the 60 that were the safest were more like trade and hands on and labor oriented. And I thought that was really interesting.
00:04:27:08 - 00:04:42:08
Speaker 1
What is it that you're seeing kind of in terms of like the biggest impact because you worked with lots of like hundreds and hundreds of startups. And while most of the 90% of the Fortune 500 and like, what do you what do you think is the case like, is that does that make sense to you or do you see something else happening?
00:04:42:10 - 00:05:11:04
Speaker 2
Yeah. it probably makes sense to me. I think if you if you believe that the direction that a lot of the AI technologies and tools are going is to essentially be able to take the place of human cognition, and therefore a lot of the roles like consultant, like lawyer and so on will be replaced. But I think if you believe that kind of, if you don't believe that, if you don't believe that's where things will ultimately go.
00:05:11:04 - 00:05:38:03
Speaker 2
And if you believe that, that the, the best place for these tools is to support humans, and to support these roles, then I don't think the roles will disappear. I think they'll be reshaped. And what success might look like in that role will be different. But also, who's successful in those roles will be different, because then it might not be a world where kind of, you know, writing the best report is what makes you a successful consultant or creating the best slide deck.
00:05:38:04 - 00:06:16:02
Speaker 2
It'll be your ability to process governance information and decide what what goes into the piece and what doesn't go into the piece, which is a very different set of skills. So so maybe if you define replacement as redefine-ment, then then then then, Yes, but but I, I, I, I struggle to see a world where you have AI essentially taking the place of humans end to end, and where they're able to invent new things and, and kind of, you know, make emotional decisions and kind of decide things, you know, for or on behalf of us.
00:06:16:02 - 00:06:18:03
Speaker 2
I struggled to see that, that future.
00:06:18:03 - 00:06:38:04
Speaker 1
Yeah. Well, it's interesting because obviously, one of the things that we all have been talking about, in modern business and, you posted something last month on, on LinkedIn that I just thought was amazing, which was a great example during the Cold War of like human in the loop. But we talk a lot in modern tech and design as well about like the human in the loop.
00:06:38:06 - 00:07:01:19
Speaker 1
It's what you're just saying, like the things that machines can't or won't replace yet at least. Right. And understanding kind of those like those unspoken social dynamics or reading between the lines or as you as you gave in that example of like using intuition mixed with expertise to like, say like override like process, right? Yeah, yeah.
00:07:01:21 - 00:07:16:18
Speaker 1
What do you think, and you've just talked about it a little bit, but like is human in the loop here to stay. And then I'm going to talk obviously your background is in largely in AI implementations and systems now. So what do you think about the human in the loop factor right now?
00:07:16:20 - 00:07:46:02
Speaker 2
I think it is here to stay, but I think I think we're at a juncture where a lot of companies have decisions to make about how they bring in and implement AI within their organizations. I think I think some organizations will take the route of, let's just bring in tools that purport to replace humans. And I do think that those companies will be in a sticky situation in six, 12, 18 months where they realize that this thing is making mistakes and this thing doesn't work as well as humans work.
00:07:46:04 - 00:08:10:04
Speaker 2
In, in every factor, And I think the companies that make purposeful, purposeful decisions about how do we thoughtfully bring in these AI tools and technologies in a way that supports our people and takes takes on the tasks that the machine can do well, but also leaves the right intentional space for human oversight? I think those are the companies that are going to thrive.
00:08:10:06 - 00:08:34:21
Speaker 1
It's interesting because your career at some point here, I really want to come back to designers, but we're going to skip it for a second and talk about the fact that your career has always spanned design and this technology. Right? Or, a technology and in your current role, both when you were at McKinsey most recently and now on your own as an advisor to both small organizations and startups as well as large enterprises on product strategy.
00:08:34:21 - 00:08:43:07
Speaker 1
But also you call it human centered AI implementation, which I think is what you're describing. But can you just describe what that is first and why it matters?
00:08:43:09 - 00:09:11:03
Speaker 2
Yeah. I think the the simple definition is kind of starting with what is the, the human need or the, the human problem you're trying to solve for and then and only then going to, okay, how do we best bring in the current suite of available technologies and how do we kind of thoughtfully bring them in at the right moments to support and augment those humans or those kind of existing workflows?
00:09:11:05 - 00:09:23:03
Speaker 2
As opposed to let's start with the AI and then work out where we can inject AI to, to kind of replace everything. And I think, yeah, I think that's, that's kind of the, the very simple definition. Yeah.
00:09:23:05 - 00:09:50:24
Speaker 1
You've been known to say that you help companies bridge the gap between AI possibility and business reality. Yeah. That sounds like a big task, especially when you're working with either, like, startup founders and their, their investors or even bigger when you're dealing with big companies. In 2025 and looking ahead from 2025, how quickly is this conversation evolving?
00:09:51:01 - 00:10:21:08
Speaker 2
Yeah. I think the I think the caveat I want to make their first is, in my experience, every business decision at its core is a human decision. You are making decisions about whether it's a group, whether it's your customers, your shareholders, your board, you're making decisions that impact and are dependent on other humans. And so when I say I help companies, you know, bridge the divide between kind of AI possibility and business reality, it's really about AI possibility and human reality.
00:10:21:10 - 00:10:41:01
Speaker 2
So I think I like to start there and I think I, and I think we are in a place right now where almost every business and I do a lot of work with large businesses, and so that's kind of that was my original domain, is struggling with the question of what do we do with AI? And they're getting pressure from shareholders.
00:10:41:06 - 00:11:05:04
Speaker 2
They're getting pressure from boards to have a concrete AI strategy to already be showing value. And so I think when you bring in that pressure with the uncertainty as to one, what's actually possible in my business or domain, because no one's really done this before. Two, you're getting a lot of marketing from AI companies saying that, you know, our tools, our technology can do everything, and it's unproven.
00:11:05:06 - 00:11:24:01
Speaker 2
And then three, there are a lot of second and third order effects that are, to a lot of businesses, yet unknown. So if I do bring in this kind of, you know, AI tool that replaces an entire part of my business, what does that mean for my other people? What does that mean in terms of the governance and oversight I need?
00:11:24:01 - 00:11:37:00
Speaker 2
What does that, how does that change my company's culture? I think that all these questions that leaders are grappling with. And so what I really try and do is, on the one hand, help them see their blind spots before they become kind of, you know,
00:11:37:02 - 00:11:39:01
Speaker 1
Flaming fires.
00:11:39:03 - 00:11:53:17
Speaker 2
Yeah. Yeah. And then to help them kind of start from the position of how do we help humans first, as opposed to how do we buy the AI technology first and then work out what to do with it?
00:11:53:19 - 00:12:20:17
Speaker 1
I mean, I imagine coming back to the topic at hand of being replaced, I imagine that that if can be communicated from an executive level or from a management suite that actually, like, really like softens the blow for people, I would imagine. Is, is that your experience that people get less nervous, they feel more like that's like it can be a partner to them, or is that not being communicated because you're still trying to explain to the management that it's possible?
00:12:20:19 - 00:12:58:23
Speaker 2
Yeah. So do you mean in terms of kind of, for example, a company bringing in an AI tool and then being able to communicate to their workforce, here is why, here are the benefits. Exactly. Yeah, and I think that's a huge factor. And I think it's a factor that a lot of companies are underestimating because I think even before generative AI, if I think back to 6 or 7 years ago when I was working on machine learning projects, we were already seeing kind of companies build a, very heavy machine learning model that might, you know, predict insurance claims or predict kind of, you know, the, the, the level to which you
00:12:58:23 - 00:13:19:24
Speaker 2
should kind of, you know, reach out to a customer early versus later on. And finding that the customer service reps who are meant to be using these models just weren't using them. And so when you have business cases predicated on 100% adoption and you have, you know, 6% adoption, you are bringing in and investing heavily in a technology that doesn't see any return.
00:13:19:24 - 00:13:45:03
Speaker 2
And and it's not until kind of, you know, you get to the point where you go, okay, why aren't people using it? Then a lot of companies go, oh, wait, we need to invest in adoption and education. And and not only that, but in starting from understanding where the people that are going to use it actually need support and how we kind of design something that works for them as opposed to designing in isolation, bringing it in and going, wait, why aren't people actually doing it?
00:13:45:03 - 00:14:18:00
Speaker 1
It's change management, right? And and organizational behavior type stuff. Yeah, yeah, yeah. It leads me to a question that, we, we originally had talked about, like what the topic of this conversation was going to be. And one of the things we had talked about, given your background, was like the future of design thinking. Design thinking seems to be this like hot button topic, but like what you've just described, I think of in terms of like the human factors in all of this, you know, is is there a role for design thinking in all of this?
00:14:18:00 - 00:14:36:10
Speaker 1
And, I don't know, maybe that's not the right terminology today, but like, what what what does that look like? Which is like, you know, the AI conversation that companies are having, as you just said, like desperately needs to explain and understand the real world consequences and its employee’s and stakeholder’s needs.
00:14:36:12 - 00:15:13:14
Speaker 2
Yeah. I yeah, absolutely. And I think there is a role I think, yeah, there's still the question of what's the right terminology, but I mean, just before this, I was speaking with a founder of a legal tech startup, and they're building an AI tool for legal professionals within companies to essentially kind of help them, I guess, speed up their processes and help them kind of make faster decisions, because a lot of kind of, a lot of procurement, a lot of contract, a lot of kind of, you know, new vendor discussions within organizations are slowed down by legal teams having to go through the process.
00:15:13:16 - 00:15:49:21
Speaker 2
Yeah. So they’re building AI that helps. But but in my experience, you know, the legal profession is a profession where there is very little new technology that has been massively adopted. And what I was saying on the call was that, you know, you have a profession that fundamentally it's success is built on reputation, risk, and trust. And if you're bringing in technology that they can't trust or they don't know how to calibrate that trust, and you haven't designed for that from the outset, you might have this technological perfect tool that has 100% accuracy, but if a lawyer goes, I'm not going to trust that,
00:15:49:21 - 00:16:08:10
Speaker 2
I'm not gonna pay money for that, then you haven't designed for the right thing. And so that's where you know, what we're talking about comes in where you start from, okay, how do we understand that? What are the right attributes they need to see in order to calibrate their trust? And how do we design from that from the beginning?
00:16:08:12 - 00:16:13:17
Speaker 2
And then that leads to decisions about how we implement the technology, what models we build and so on.
00:16:13:19 - 00:16:44:19
Speaker 1
It's totally interesting. And it leads me, I mean, you've obviously, as we've talked about, you know, your background and you've worked in lots of different industries. But I think some of the ones that I like jumped out at me as we, as I was looking at your background is like retail, B2B enterprise, banking and finance, as you and you just mentioned, legal. Like, are there specific sectors in your opinion where this transformation, let's call it digital transformation, you know, v5.0 at this point or something is most relevant?
00:16:44:19 - 00:16:50:08
Speaker 1
And where there might be this biggest risk of like coming back to our topic of like being replaced.
00:16:50:10 - 00:17:23:21
Speaker 2
Yeah, I think any industry or profession where, where the, the tools and mechanisms that that humans use to be great at their jobs is written down. So I think that's why I like computer programing and coding has there's been massive impact because, you know, code is is absolute. There is no inference. There's no kind of, there's nothing kind of, you know, spoken in kind of words or body language.
00:17:23:21 - 00:17:41:07
Speaker 2
It is the code you write and, and the the pull request and the notes you write against it are there. So when you have a large language model that learns from text, then, yeah, it can learn to be a fantastic coder very well because 100% of what it needs is there and it's written down. So that's an example.
00:17:41:10 - 00:18:12:02
Speaker 2
But I think if you have, you know, business relationships, or if you have kind of, you know, you know, running a shop where success is not only dependent on are you functionally excellent, but can you, you know, do people want to be around you? Do people trust you? Those, aren't written down and so so I think I think the industries and professions where there's, where things are heavily codified, written down, are probably going to be at biggest threat.
00:18:12:04 - 00:18:19:10
Speaker 2
And, and I think everything else that still requires human relationships is, is going to be fine, at least for a while. Yeah.
00:18:19:12 - 00:18:39:24
Speaker 1
It makes total sense. I want to come back to, like, your expertise in design, Tey, because it kind of is this, like, red thread throughout all of it. I get the feeling in my role at iF Design, you know, when I speak to designers and design organizations, I don't get the sense that designers themselves fear that they're going to become obsolete.
00:18:40:05 - 00:19:02:20
Speaker 1
But I get the sense that they fear that their organizations are going to see them as obsolete, which is. Yeah, it's it is an interesting kind of, line to fine line to walk, right. Talk about this a little bit. I mean, like and, and do you see the same thing or what do you see?
00:19:02:22 - 00:19:26:16
Speaker 2
Yeah, I, I think it's a valid fear. I do think that a lot of organizations, if not already, are going to see design as obsolete because I think a lot of the of, of leaders in organizations have yet to actually tangibly see or measure the value of good design, and therefore they see design as the surface level, what it looks like.
00:19:26:16 - 00:20:04:18
Speaker 2
And then when you see tools that can create kind of really great looking user interfaces, you go, okay, why do I need all these people? So I, I am very much looking forward to seeing companies fire their entire design teams, and then six months later, be in the news because they suck. Excuse my French. So I yeah, I think there will be many cautionary tales because I think a lot of the value of great designers is not in the polished work they produce, but the thinking and, the judgment and the taste that goes into that work.
00:20:04:20 - 00:20:17:11
Speaker 2
And, and being able to communicate with stakeholders and understand intent, things that kind of, you know, that that kind of using a visual AI tool aren't going to replicate or replace. Yeah.
00:20:17:13 - 00:20:45:17
Speaker 1
That it that is a fact. It's interesting. You wrote a book not so not so long ago, or co-wrote a book, I think, called “The AI Revolution”. Congrats first of all. Thank you. But in it, you argue that AI is fundamentally, fundamentally breaking our relationship with failure and learning. When I asked you to define being replaced, or in the very beginning, you kind of had two pathways.
00:20:45:23 - 00:20:54:07
Speaker 1
And the second one, I think is this, right? Talk about what you mean. And also where could that lead humankind?
00:20:54:09 - 00:21:33:12
Speaker 2
Yeah. So I think that the root of that is, I mean, there's a lot of brilliant research around how humans learn, both from a very young age and, and how adults learn. And a big part of that learning comes from making mistakes. And, and, you know, there's a lesson in every failure. And if, if, if you now have tools and technologies that sound very confident, give you answers straight away and don't allow you to to embark on that process of learning, making little mistakes, self-correcting, adjusting, developing that part of your mind that allows you to have that intuition to go, okay, is this right or wrong?
00:21:33:14 - 00:21:43:04
Speaker 2
I think that's that's where I was coming from in terms of that fear. And we are already seeing research. I think there was a piece from, from MIT or or Harvard a few months ago.
00:21:43:06 - 00:21:43:18
Speaker 1
MIT.
00:21:43:21 - 00:21:51:10
Speaker 2
That, yeah, that showed that people who regularly use tools like ChatGPT just to get answers are already losing some of that cognition.
00:21:51:12 - 00:22:03:20
Speaker 1
Absolutely. It's and or I think they also gave the example of, like writing an essay, like they didn't recall anything about what they had just like submitted, even though they had been working on it. So that's pretty wild.
00:22:03:22 - 00:22:05:19
Speaker 2
Yeah. Yeah, yeah.
00:22:05:21 - 00:22:21:02
Speaker 1
What separates real technology or technology that solves real problems, from technology that either is just evolutionary or even worse, that just creates new problems.
00:22:21:04 - 00:22:44:18
Speaker 2
Yeah, honestly, it's the people that create these technologies because I think, I think it's easy to see AI or ChatGPT as this thing that just exists and just gives you answers. But every facet of that was was a decision that a set of humans made about how to build and design this, which models to use, which data was used to train this.
00:22:44:18 - 00:23:08:03
Speaker 2
And so I do think that kind of intentional design that solves real problems comes from people that start the process with, let's really deeply look at the problem and understand if it is a problem worth solving, and then build everything from there, as opposed to starting with the technology and then then kind of somewhere down the line going, okay, where, where, where can we use this?
00:23:08:03 - 00:23:32:14
Speaker 2
Because typically by then you've already lost. If you're, if you, if you, if you don't have a clear problem you're solving and it might be in the short term that you don't get adoption, it might be in the longer term when people realize that there are very negative second and third order effects. But but, but I think, I think that the root is the people making decisions and the intention behind those decisions.
00:23:32:16 - 00:24:01:19
Speaker 1
It always, it's always it's always the people at the at the end of the day. I, I have one last question before we get to our final closing question, Tey. Obviously right now, you had said like seven years ago, it was machine learning. In 2025, it's like all about AI, right, among the emerging technologies. Are there other things in this, let's call it emerging tech space that are potentially happening that you think either will impact and or influence this concept of being replaced?
00:24:01:21 - 00:24:15:13
Speaker 2
Yeah. I think vision models aren't talked about enough. So, so kind of, I spend a little bit of time, I've spent a little bit of time over the last few weeks with Meta's Ray-Ban glasses, and
00:24:15:13 - 00:24:17:14
Speaker 1
iF Design award winner 2024.
00:24:17:19 - 00:24:44:02
Speaker 2
Yeah, yeah. And the ability to have technology that not only sees what you're seeing, but is able to make almost immediate calculations. So like, I saw a demo where like, it was, you know, someone walking through a city and you're able to have like immediate recall of how many red shirts that I walked past, or, you know, how many females aged 35 to 75?
00:24:44:04 - 00:25:10:08
Speaker 2
you know, did I walk past between 7 p.m. and 7:06 p.m.? So, so I mean, that's a very simple example, but when you have technology that's able to kind of give you instant recall, I think that opens up a lot of use cases around kind of surgery, around kind of, you know, repair, you know, imagine being able to kind of repair a washing machine just with kind of, you know, something that's giving you direct instructions and able to see what you're seeing.
00:25:10:08 - 00:25:23:11
Speaker 2
So I do think there are a lot of useful and valuable problems to solve that help real people that aren't being talked about that are using some of these technologies that aren't as kind of cool as ChatGPT, for example.
00:25:23:13 - 00:25:43:10
Speaker 1
Really interesting. So, second to last question, Tey, is there besides your book, obviously, is there like any like clear reference or tool or place that you would like to send our viewers and listeners if they want to understand this topic more?
00:25:43:12 - 00:26:07:06
Speaker 2
Yes. So this book, it's called “Co-Intelligence” by Ethan Mollick, fantastic. Fantastic book that that talks a lot and looks into the future of work and how AI tools and technologies could impact that. And also talks a lot about what you need to start thinking about now if you really want to understand these tools in the right way.
00:26:07:06 - 00:26:32:16
Speaker 2
So that's one thing I definitely recommend. There's also like a fantastic series, I believe it's on YouTube by Andrew Ng, and it's just like an introduction to AI and generative AI, and he talks about a lot of the core principles, just in a super accessible way. So that's something I recommend to not just designers, but anyone to just better understand how generative AI, large language models how they work.
00:26:32:16 - 00:26:43:00
Speaker 2
Because once you understand how they work, you tend to better understand how to trust them and how to to to work out what to use them for and what not to use them for.
00:26:43:02 - 00:26:59:23
Speaker 1
Thank you for that. That's like I hadn't even prepped you for that question. So thanks for being on hand, that's amazing. Last question is the same as we always ask, every guest, what's your greatest hope for the future of being replaced in 25 years? So like 2050?
00:27:00:00 - 00:27:37:00
Speaker 2
Yeah. 2050. My biggest hope is that being replaced means that the tasks that are not well suited to humans doing and that add stress and friction rather than bring joy, are replaced. And we're in a world where where humans, the planet, the world we live in is better off as a result of the technologies that we're bringing in, because they've been thoughtfully and intentionally designed to solve real problems.
00:27:37:02 - 00:27:42:05
Speaker 2
And, and also a world where everything isn't just about making money. I just wanted to throw that in there.
00:27:42:05 - 00:27:52:00
Speaker 1
Oh, I'll go with you on that one. Tey Bannerman, thanks for joining us on Future of XYZ today, the day before Halloween live from London.
00:27:52:02 - 00:27:54:02
Speaker 2
Thank you so much for having me.
00:27:54:04 - 00:28:13:14
Speaker 1
All of our watchers and listeners, if, you didn't already know, you can get all show notes, links, etc. at ifdesign.com/futureofxyz. You can find us anywhere you get your favorite podcasts, please leave us a five star review so others can find us. You can watch on YouTube or follow on Instagram.
00:28:13:14 - 00:28:18:10
Speaker 1
Thank you again, Tey, and we will see you all again in two weeks.
00:28:18:12 - 00:28:19:09
Speaker 2
Thank you.